DXVK Is Aiming To Get Direct3D 11 Over Vulkan In Wine

[phoronix]

Phoronix: DXVK Is Aiming To Get Direct3D 11 Over Vulkan In Wine

While there's long been VK9 as an effort to get Direct3D 9 running atop Vulkan and Wine developers planning VKD3D for Direct3D 12 over Vulkan, now coming in between is DXVK for Direct3D 11 support...

http://www.phoronix.com/scan.php?page=news_item&px=DXVK-Direct3D-11-Vulkan

[ssorgatem]

What is the point of this project?
Is better performance expected by using Vulkan as a backend instead of OpenGL?

[boxie]

What is the point of this project?
Is better performance expected by using Vulkan as a backend instead of OpenGL?

yes, because vulkan is a lower level, higher level things can be implemented on top of it. it won't be as fast as a native implementation, but it should do the trick.

[boxie]

also - VK9 is out with another blog post Michael http://techtinker.blogspot.com.au/20...eport-nov.html

[R41N3R]

Personally I'm very excited about VK9, VKD3D and now DXVK :-) My personal experience with Wine is that a lot of games don't work and while there are e.g. wine-d3d9 patches to get a better performance, it doesn't benefit the whole platform. Running everything on top of Vulkan sounds much more reasonably to me and it will hopefully reduce bugs and increase performance.

[L_A_G]

One thing that has always worried me about VK9, and which applies to this as well, is the difficulty of debugging and the potential effect this can have on the final product. With an explicit "no-handholding" type API like Vulkan there's obviously just so much more that can go wrong and when things do go wrong you're not necessarily going to get any of the detailed logs the always-active support code in OpenGL and D3D provides. Specially end users are much less able to assist in debugging stuff when they can only really provide details on how they got the application to crash and not any of the detailed logs they'd have to offer on OpenGL and D3D.

[geearf]

Isn't Direct3D 11 an extension of Direct3D 10? If so, wouldn't it make sense for it to first support 10?

[VikingGe]

Is better performance expected by using Vulkan as a backend instead of OpenGL?

Developer here. The whole point is to cut down the CPU overhead compared to wined3d. D3D11->OpenGL is not cheap, and OpenGL itself is not cheap either. In fact, if operf doesn't deceive me, Wine games tend to spend more time in wined3d.dll than in the actual OpenGL driver (at least on Mesa), and pretty much all D3D11 games i've played on Wine so far are heavily CPU-bound.

OpenGL also forces you to serialize pretty much everything into one single thread whereas D3D11 is kind of thread-aware in that it decouples resource creation (ID3D11Device) from the actual rendering commands (ID3D11DeviceContext), which can be exploited in Vulkan without any sort of voodoo magic. Not to mention that the underlying driver is a lot thinner.

Isn't Direct3D 11 and extension of Direct3D 10? If so, wouldn't it make sense for it to first support 10?

Yes and no. Hardly anyone uses pure D3D10, but it might actually be possible to implement 10 on top of 11 (as it is mostly a subset) or even to implement both D3D10 and D3D11 COM interfaces in the same classes. But right now there are no such plans.

[artivision]

I hope DXVK can play games inside 2018.

[starshipeleven]

With an explicit "no-handholding" type API like Vulkan

Vulkan is lower-level, not "no-handholding". There are debug tools for that too.